Enhancing thin-pay estimation through stochastic simultaneous inversion

Purnomo, Eko (Universiti Teknologi PETRONAS) | Ghosh, Deva (Universiti Teknologi PETRONAS)


A new approach in Bayesian stochastic seismic inversion has been developed to accurately invert thin pay bed reservoirs below “so called” Seismic Resolution. Sensitive thin bed “attributes”, e.g. wavelet, tuning and spectral analysis, well log, and locally varying anisotropy (LVA), are applied to constrain the proposed Bayesian seismic inversion. The prior model is generated through a perturbation of apparent reflectivity from seismic data. This process guides the perturbation to capture as much as possible detail information of the properties to be inverted. The LVA is incorporated to impose the spatial continuity of the inverted parameters. The low frequency model is built from well log data, and included iteratively to the prior model through a process of frequency matching. The misfit between modelled and observed data is controlled by an energy “spectral attribute” to ensure better resolution. Finally, Markov Chain Monte Carlo method is employed to conduct the simulation. Therefore the minimum inversion biasness is ensured and better uncertainty assessment is provided. This proposed stochastic inversion is applied to an offshore field in Malaysia and successfully inverts thin 5 m gas sand buried at the depth of 1100 m. This thin reservoir was identified on well log but was not detected a) either on the high resolution 3D seismic or b) after deterministic Simultaneous Inversion. This breakthrough will open a new avenue while exploring and developing thin stacked pays very common in the Malaysian and South East Asia offshore Basins. Constraints and high quality in modern seismic data ensure our success.

Presentation Date: Wednesday, September 27, 2017

Start Time: 9:20 AM

Location: 370D

Presentation Type: ORAL